2GB File Size Limit in Linux??

[Armitage]

Sure back in the bad old 2.2 days! But not with a 2.4.18 kernel!

I've got a simple C++ sim going that generates a ton of data. When the ASCII output file hit the 2GB mark, it died with a message "File size limit exceeded".
What's wierd is that I can use dd to create a file > 2GB on the same partition, so it seems like it must be a C++ limitation, not a filesystem limitation. But I know I've used C/C++ programs to generate > 2GB files in the past (on a different machine, different kerner, distro, etc.). Any ideas??

Relavent information:
RedHat 8.0 w/ 2.4.18
ext3 on software RAID1
gcc v3.2.7

I'm writing to a standard fstream object.

 

[Armitage]

ugg ... gcc bug

Linkage

Kind of disturbing to see it marked "critical" but unassigned after > 1 month. It seems to be a relatively recent bug and also doesn't affect stdio which explains why my other program worked (didn't use iostream).

 

[Armitage]

Aha, its not assigned because its a duplicate of this, where work seems to have stalled early in Feb. :brokenheart:

Don't mind me ... just talking to myself here.

 

[igowerf]

 

[kt]

I'll chime in so you don't feel like you've been talking to yourself..

 

[Armitage]

GCC 3.3 released today ... still no fix

 

[n0cmonkey]

Any clue if this is a bug with better version of gcc or is it just the 3.x series?

 

[Armitage]

I know for a fact that it worked with RedHat GCC from 7.2 (AKA 2.96) and with the 2.95 versions. I don't know about the 3.0, I didn't start using 3.* until 3.1

 

[n0cmonkey]

Originally posted by: ergeorge
I know for a fact that it worked with RedHat GCC from 7.2 (AKA 2.96) and with the 2.95 versions. I don't know about the 3.0, I didn't start using 3.* until 3.1

Well, maybe there is a way to be able to use both 2.95 and 3.x on the same system.

 

[Armitage]

Yea, I've done that before ... don't recall why. It's something of a PITA
For now, I've changed the io in the affected app from iostream to stdio ... which was easy for this one because it was a trivial program which just happened to generate an obscene amount of output
But another of our major programs may be coming online again soon, and I know for certain this breaks it , although there may be some workarounds.

If my PR gets approved, we'll be getting a new xeon based cluster soon, and the intel compiler suite. I'll have to see if the intel compilers use the gcc libstdc++, or provide their own.

edit: Actually, RedHat 7.* shipped with both a 2.9* GCC and a 3.* GCC ... so it's definitely possible to have more then one on a system.

 

[Barnaby W. Füi]

Originally posted by: n0cmonkey
Any clue if this is a bug with better version of gcc or is it just the 3.x series?

What would the better version be?

 

[Armitage]

Alot of people haven't moved to the 3.* releases yet. For one thing, it breaks alot of broken code that older versions accepted, and early releases were VERY slow compiling.
I was happy with it up until this mess.

 

[n0cmonkey]

Originally posted by: BingBongWongFooey

Originally posted by: n0cmonkey
Any clue if this is a bug with better version of gcc or is it just the 3.x series?

What would the better version be?

A version that works for what you need to do with it

 

[Armitage]

update ... target to fix this is GCC 3.4 "sometime this year"

 

[ProviaFan]

Originally posted by: ergeorge
update ... target to fix this is GCC 3.4 "sometime this year"

Heh, that seems like a pretty serious bug; I wish they would fix it soon, but the chances of any Open-Source software coming out "on time" are virtually nil. :disgust:

<-- is going to learn C++ and this bug would probably affect a program he will be writing. :frown: